A5069991677- Molecular Biology Techniques and Applications
- Forensic and Genetic Research
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- MicroRNA in disease regulation
- Advanced Neural Network Applications
- Cancer-related molecular mechanisms research
- Metabolomics and Mass Spectrometry Studies
- Epigenetics and DNA Methylation
- Corneal surgery and disorders
- Sperm and Testicular Function
- Genomics and Phylogenetic Studies
- COVID-19 diagnosis using AI
- Video Surveillance and Tracking Methods
- Human Pose and Action Recognition
- Retinal Diseases and Treatments
- Forensic Entomology and Diptera Studies
- Glaucoma and retinal disorders
- Retinopathy of Prematurity Studies
- E-commerce and Technology Innovations
- Genetic Associations and Epidemiology
- Complex Network Analysis Techniques
- Peanut Plant Research Studies
- Visual Attention and Saliency Detection
- Gait Recognition and Analysis
Singapore University of Technology and Design
2023-2025
Wenzhou Medical University
2011-2025
Suzhou Municipal Hospital
2018-2025
Nanjing Medical University
2024-2025
China Agricultural University
2022-2025
Affiliated Eye Hospital of Wenzhou Medical College
2025
Guangdong Medical College
2024
West China Medical Center of Sichuan University
2017-2022
Sichuan University
2015-2022
Chinese Academy of Agricultural Sciences
2021
Deep models trained on source domain lack generalization when evaluated unseen target domains with different data distributions. The problem becomes even more pro-nounced we have no access to samples for adaptation. In this paper, address generalized semantic segmentation, where a segmentation model is be domain-invariant without using any data. Existing approaches tackle standardize into unified distribution. We argue that while such standardization promotes global normalization, the...
Myopic choroidal neovascularization (CNV) is a common reason for visual impairment. This study investigated the clinical effects of repeated intravitreal injections ranibizumab among patients with CNV secondary to pathologic myopia. involved single-center, non-randomized prospective cohort research design including 39 myopic and control group 10 cataract. Plasma aqueous humor samples were analyzed compare cytokine concentrations between two groups assess changes after injections....
Semantic segmentation is a crucial image understanding task, where each pixel of categorized into corresponding label. Since the pixel-wise labeling for ground-truth tedious and labor intensive, in practical applications, many works exploit synthetic images to train model real-word semantic segmentation, i.e., Synthetic-to-Real Segmentation (SRSS). However, Deep Convolutional Neural Networks (CNNs) trained on source data may not generalize well target real-world data. To address this...
Domain adaptation is critical for success when confronting with the lack of annotations in a new domain. As huge time consumption labeling process on 3D point cloud, domain semantic segmentation great expectation. With rise multi-modal datasets, large amount 2D images are accessible besides clouds. In light this, we propose to further leverage data by intra and inter cross modal learning. intra-domain learning, most existing works sample dense pixel-wise features into same size sparse...
Translating images from a source domain to target for learning models is one of the most common strategies in adaptive semantic segmentation (DASS). However, existing methods still struggle preserve semantically-consistent local details between original and translated images. In this work, we present an innovative approach that addresses challenge by using sourcedomain labels as explicit guidance during image translation. Concretely, formulate cross-domain translation denoising diffusion...
Unsupervised Domain Adaptation (UDA) is quite challenging due to the large distribution discrepancy between source domain and target domain. Inspired by diffusion models which have strong capability gradually convert data distributions across a gap, we consider explore technique handle UDA task. However, using different domains non-trivial problem as standard generally perform conversion from Gaussian instead of specific distribution. Besides, during conversion, semantics source-domain needs...
AIM: To investigate the refractive and histological changes in guinea pig eyes after posterior scleral reinforcement with allografts. METHODS: Four-week-old pigs were implanted allografts, of refraction, corneal curvature axis length monitored for 51d. The effects methylprednisolone (MPS) on refraction parameters also evaluated. And microstructure ultra-microstructure observed 9d 51d operation. Repeated-measures analysis variance one-way used. RESULTS: outcome eye decreased operation, change...
Text-to-Image (T2I) models have advanced significantly, but their growing popularity raises security concerns due to potential generate harmful images. To address these issues, we propose UPAM, a novel framework evaluate the robustness of T2I from an attack perspective. Unlike prior methods that focus solely on textual defenses, UPAM unifies both and visual defenses. Additionally, it enables gradient-based optimization, overcoming reliance enumeration for improved efficiency effectiveness....
We propose a weakly supervised approach for salient object detection from multi-modal RGB-D data. Our only relies on labels scribbles, which are much easier to annotate, compared with dense used in conventional fully setting. In contrast existing methods that employ supervision signals the output space, our design regularizes intermediate latent space enhance discrimination between and non-salient objects. further introduce contour branch implicitly constrain semantic boundaries achieve...